Since genetic approaches are not an option to chart the connectivity of the human brain how could we make progress here? As the human brain is comparatively large we have the advantage of being able to use non-invasive imaging techniques to obtain structural and functional images with a resolution in the millimeter range. As defined above this would allow us to detect regional connectivity but not neuronal connectivity, and at present we have to rely on indirect measurements indicating anatomical connections. Creating the connectivity matrix of the human brain, however, is a worthwhile endeavour with great importance for cognitive neuroscience and neuropsychology. It would form a unique neuroinformatics resource for reference in a variety of fields, which would thereby make closer contact and maybe allow cross-linking of studies that were conceived with a very specific question in mind but have wider implications.
As we suggested previously (Sporns et al. 2005) assembling the connectivity matrix of the human brain would require a multi-modal approach, starting from the macroscopic level using structural and functional in vivo imaging techniques. It would eventually make contact with the mesoscopic level of cortical columns and layers, information that is conceived to be generic to various locations and would thus be mapped from a certain zoom level onwards independent of the exact coordinates. Cellular information (including animal to the mesoscopic level data) could then be linked to maintaining strict transparency of the origin of the data. In pursuing this goal a series of steps would have to be performed, which build on the issues faced with validating DTI as predicting anatomical connections (see section on Regional Connectivity above).
• Step 1: Probabilistic tractography of diffusion-weighted imaging data starting with thalamocortical tracts followed by U-fibres of all cortical regions ultimately resulting in a voxel-wise probabilistic all-to-all structural connectivity matrix.
• Step 2: Correlation analysis of spatially registered and equally resolved resting activity or multi-stimulus/multi-task activity data (fMRI and/or
MEG) in the same person resulting in a voxel-wise all-to-all functional connectivity matrix.
• Step 3: Cluster comparison between the structural and the functional connectivity matrix identifying regions of consistent structure-function relationships.
• Step 4: Comparison of human analyses (step 3) with structural and functional macaque data to identify correspondences (e.g. visuo-motor pathways) and deviations (speech: fasciculus arcuatus?).
• Step 5: Validation of strongest predictions from final connectivity matrix using custom-designed stimuli and transcranial stimulation in combination with behavioural testing and functional imaging.
• Step 6: Population analysis of healthy subjects and spatial registration to standard brain for probabilistic statements about data from steps 1-5.
• Step 7: Comparison of population data on clustered brain regions to histologically identified regions in probabilistic human brain atlas to assess correspondence.
• Step 8: Comparison of population data between healthy subjects and patient groups supposedly suffering from connectional disorder (e.g., white matter stroke, multiple sclerosis, chronic schizophrenia) using same tasks with similar performance measures.
The steps described here are clearly not the end of this project. However, they are a beginning and give an idea of the opportunities and the challenges that lie ahead of us if we pursue this direction. Even before we tackle the task of relating mesoscopic data to this macroscopic framework, there are enough detailed problems (e.g. concerning spatial normalization of data sets within and across individuals, as well as the homology issue that is often ignored when comparing across species) that need to be addressed.
Was this article helpful?